Attribution Non-Commercial (BY-NC)

11 tayangan

Attribution Non-Commercial (BY-NC)

- Biogeography Based Optimization Approach for Optimal Power Flow Problem Considering Valve Loading Effects
- Modeling,simulation and control of a drum boiler
- Nesting
- LIne Balancing
- Umbau1
- orbit
- J3 - Elsevier JOCS - a New Population Seeding Technique for Permutation-Coded Genetic Algorithm
- Action Selection and Obstacle Avoidance Using Ultrasonic and Infrared Sensors
- IJERA 012165172S
- Blind Source Separation of Super and Sub-Gaussian Signals with ABC Algorithm
- Polymer Extrusion
- dissertation
- GeneticAlgorithm Tutorial
- A “quasi-genetic” algorithm for searching the dangerous areas generated by a grounding system
- 86-199_Pages-5
- ijicic-10-01002
- 06198104
- breeder93
- CLSP(Miltiple Supplier)
- SJEAT 2(11)419-427

Anda di halaman 1dari 14

William H. Hsu

Department of Computing and Information Sciences

Kansas State University

234 Nichols Hall

Manhattan, KS 66506-2302

USA

voice: +1 785-532-6350

fax: +1 785-532-7353

email: bhsu@cis.ksu.edu

Genetic Algorithms

INTRODUCTION

genetic programming (GP), they are one of the main classes of genetic and evolutionary

(Goldberg, 1989; Wikipedia, 2004). Traditionally, solutions are represented using fixed-

length strings, especially binary strings, but alternative encodings have been developed.

according to some probability distribution, usually uniform and updates this population in

steps called generations. Each generation, multiple individuals are randomly selected

from the current population based upon some application of fitness, bred using crossover,

problem

This chapter begins with a survey of GA variants: the simple genetic algorithm,

clustering, and feature selection and extraction. It concludes with a discussion of current

issues in GA systems, particularly alternative search techniques and the role of building

BACKGROUND

The field of genetic and evolutionary computation (GEC) was first explored by

Turing, who suggested an early template for the genetic algorithm. Holland performed

much of the foundational work in GEC in the 1960s and 1970s. His goal of

artificial systems led to the formulation of the simple genetic algorithm (Holland, 1975).

State of the field: To date, GAs have been successfully applied to many significant

problems in machine learning and data mining, most notably classification, pattern

detectors (Rizki et al., 2002; González & Dasgupta, 2003) and predictors (Au et al.,

Theory of GAs: Current GA theory consists of two main approaches – Markov chain

analysis and schema theory. Markov chain analysis is primarily concerned with

characterizing the stochastic dynamics of a GA system, i.e., the behavior of the random

sampling mechanism of a GA over time. The most severe limitation of this approach is

that while crossover is easy to implement, its dynamics are difficult to describe

mathematically. Markov chain analysis of simple GAs has therefore been more

(Schwefel, 1977).

slow down processing and can also resulting in a phenomenon called takeover where the

state innovation” in GAs as the situation where time to produce a new, more highly-fit

building block (the innovation time, ti) is lower than the expected time for the most fit

individual to dominate the entire population (the takeover time, t*). “Steady state

t*, because the “countdown” to takeover or “race” between takeover and innovation is

reset.

MAIN THRUST OF THE CHAPTER

The general strengths of genetic algorithms lie in their ability to explore the search

space efficiently through parallel evaluation of fitness (Cantú-Paz, 2000) and mixing of

partial solutions through crossover (Goldberg, 2002); maintain a search frontier to seek

global optima (Goldberg, 1989); and solve multi-criterion optimization problems. The

basic units of partial solutions are referred to in the literature as building blocks or

schemata. Modern GEC systems are also able to produce solutions of variable length

learning problems. Examples of this include induction of decision trees (Cantú-Paz &

Types of GAs

reals. The simple GA performs crossover and mutation at the bit level for all of these.

Other variants treat the chromosome as a parameter list, containing indices into an

instruction table or an arbitrary data structure with pre-defined semantics, e.g., nodes in a

linked list, hashes, or objects. Crossover and mutation are required to preserve semantics

by respecting object boundaries, and formal invariants for each generation can specified

according to these semantics. For most data types, operators can be specialized, with

Applications

et al. produced GABIL (Genetic Algorithm-Based Inductive Learning), one of the first

general-purpose GAs for learning disjunctive normal form concepts. (De Jong et al.,

1993). GABIL was shown to produce rules achieving validation set accuracy comparable

Since GABIL, there has been work on inducing rules (Zhou et al., 2003) and

decision trees (Cantú-Paz & Kamath, 2003) using evolutionary algorithms. Other

representations that can be evolved using a genetic algorithm include predictors (Au et

al., 2003) and anomaly detectors (González & Dasgupta, 2003). Unsupervised learning

methodologies such as data clustering (Hall et al., 1999; Lorena & Furtado, 2001) also

admit GA-based representation, with application to such current data mining problems as

gene expression profiling in the domain of computational biology (Iba, 2004). KDD

from text corpora is another area where evolutionary algorithms have been applied

features (Raymer et al., 2000), selecting features (Hsu, 2003), or selecting training

instances (Cano et al., 2003). They have also been applied to combine, or fuse,

FUTURE TRENDS

Some limitations of GAs are that in certain situations, they are overkill compared to

neural networks using backpropagation, and even simulated annealing and deterministic

global search. In global optimization scenarios, GAs often manifest their strengths:

efficient, parallelizable search; the ability to evolve solutions with multiple objective

criteria (Llorà & Goldberg, 2003); and a characterizable and controllable process of

innovation.

Opinion is, however, divided over the importance of crossover verses mutation.

Some argue that crossover is the most important, while mutation is only necessary

to ensure that potential solutions are not lost. Others argue that crossover in a

• In the field of GEC, basic building blocks for solutions to engineering problems

have primarily been characterized using schema theory, which has been critiqued

a GA. Proponents of schema theory have shown that it provides useful normative

Recent and current research in GEC relates certain evolutionary algorithms to ant colony

CONCLUSION

learning and optimization. It has been shown to be efficient and powerful through many

The current literature (Goldberg, 2002; Wikipedia, 2004) contains several general

• GAs are sensitive to deceptivity, the irregularity of the fitness landscape. This

includes locally optimal solutions that are not globally optimal; lack of fitness

criteria. This phenomenon, called concept drift in supervised learning and data

• GAs are not always effective at finding globally optimal solutions, but can rapidly

locate good solutions, even for difficult search spaces. This makes steady-state

GAs (Bayesian optimization GAs that collect and integrate solution outputs after

to generational GAs (maximization GAs that seek the best individual of the final

Looking ahead to future opportunities and challenges in data mining, genetic

also provide a practical method for optimization of data preparation and data

transformation steps. The latter includes clustering, feature selection and extraction,

instance selection. In data mining, GAs are likely to be most useful where high-level,

adaptive step size) and multi-objective data mining are also problem areas where GAs

REFERENCES

Atkinson-Abutridy, J., Mellish, C., & Aitken, S. (2003). A semantically guided and

Au, W.-H., Chan, K.C.C., & Yao, X. (2003). A novel evolutionary data mining

Cano, J.R., Herrera, F., & Lozano, M. (2003). Using evolutionary algorithms as instance

MA: Kluwer.

Cantú-Paz, E. & Kamath, C. (2003). Inducing oblique decision trees with evolutionary

De Jong, K.A., Spears, W.M., & Gordon, F.D. (1993). Using genetic algorithms for

Goldberg, D.E. (2002). The Design of Innovation: Lessons from and for Competent

González, F.A. & Dasgupta, D. (2003). Anomaly Detection Using Real-Valued Negative

Hall, L.O., Ozyurt, I.B., & Bezdek, J.C. (1999). Clustering with a genetically optimized

Holland, J.H. (1975). Adaptation in natural and artificial systems, Ann Arbor: The

Evolutionary Computation (Banzhaf, W. & Foster, J., guest editors), 5(2), 145-

156.

Kargupta, H. & Ghosh, S. (2002). Toward Machine Learning Through Genetic Code-

258.

Kuncheva, L.I., Jain, L.C. (2000). Designing classifier fusion systems by genetic

Lorena, L.A.N. & Furtado, J. C. (2001). Constructive Genetic Algorithm for Clustering

Parpinelli, R.S., Lopes, H.S., & Freitas, A.A. (2002). Data mining with an ant colony

6(4):321- 332.

Raymer, M.L., Punch, W.F., Goodman, E.D., Kuhn, L.A., Jain, A.K. (2000).

Rizki, M.M., Zmuda, M.A., & Tamburino, L.A. (2002). Evolving pattern recognition

Birkhauser Verlag.

http://en.wikipedia.org/wiki/Genetic_algorithm.

Zhou, C., Xiao, W., Tirpak, T. M., & Nelson, P.C. (2003). Evolving accurate and

TERMS AND DEFINITIONS

are paired up and exchange some portion of their genetic sequence. Crossover in GAs is

highly stylized and typically involves exchange of strings. These can be performed using

a crossover bit mask in bit-string GAs, but require complex exchanges (such as partial-

selection, which begins with a set of potential solutions, then iteratively applies

algorithms to generate new candidates and select the fittest from this set. The process

Generation: The basic unit of progress in genetic and evolutionary computation, a step

in which selection is applied over a population. Usually, crossover and mutation are

typically represented using strings (often of fixed length) and permutations in genetic

algorithm around an inducer to find locally optimal parameter values such as the relevant

feature subset for a given classification target and data set. Validation set accuracy is

typically used as fitness. Genetic algorithms have been used to implement such wrappers

candidate solution. These include bit-reversal in bit-string GAs and shuffle and swap

Problem). Permutation GAs use specialized crossover and mutation operators compared

corresponding to a set of individuals. Schemata are typically denoted by bit strings with

instances, one for each instantiation of the # symbols to 0 or 1. Schemas are important in

GA research because they form the basis of an analytical approach called schema theory,

for characterizing building blocks and predicting their proliferation and survival

reproduce, and the basis of speciation according to the Darwinian theory of evolution.

cases, with the combined evaluation measures being compared in order to choose

individuals.

i

Payoff-driven reinforcement learning describes a class of learning problems for intelligent agents that

receive rewards, or reinforcements, from the environment in response to actions selected by a policy

function. These rewards are transmitted in the form of payoffs, sometimes strictly nonnegative. A GA

acquires policies by evolving individuals, such as condition-action rules, that represent candidate policies.

- Biogeography Based Optimization Approach for Optimal Power Flow Problem Considering Valve Loading EffectsDiunggah olehIDES
- Modeling,simulation and control of a drum boilerDiunggah olehDipankar Kalita
- NestingDiunggah olehMahmoud Reda ElSherif
- LIne BalancingDiunggah olehManish Bachlaus
- Umbau1Diunggah olehLiviu Badea
- orbitDiunggah olehPraveen Kumar Aradhyula
- J3 - Elsevier JOCS - a New Population Seeding Technique for Permutation-Coded Genetic AlgorithmDiunggah olehRajakumar Ramalingam
- Action Selection and Obstacle Avoidance Using Ultrasonic and Infrared SensorsDiunggah olehJoy del Rosario
- IJERA 012165172SDiunggah olehchetan2042
- Blind Source Separation of Super and Sub-Gaussian Signals with ABC AlgorithmDiunggah olehIDES
- Polymer ExtrusionDiunggah oleheduardo_um
- dissertationDiunggah olehapi-286312075
- GeneticAlgorithm TutorialDiunggah olehNermeen Kamel
- A “quasi-genetic” algorithm for searching the dangerous areas generated by a grounding systemDiunggah olehbritsnj
- 86-199_Pages-5Diunggah olehZellagui Energy
- ijicic-10-01002Diunggah olehHanan Mouchtakiri
- 06198104Diunggah olehNorbey Marin Moreno
- breeder93Diunggah olehDeepak Kumar Singh
- CLSP(Miltiple Supplier)Diunggah olehHadeel Fayez
- SJEAT 2(11)419-427Diunggah olehsmprabu24317
- CrlMultiareaEPSDiunggah olehJasiel Zurita Santiago
- AIAA-1995-1218-483.pdfDiunggah olehrs0004
- 17 ConclusionDiunggah olehmhk665
- TD06_002Diunggah olehapi-26622648
- A Generalized Pattern Matching Approach for Multi-stepDiunggah olehKent Choo
- IIM_2013052413184014Diunggah olehAbouSalsabil
- 10.1.1.31.316Diunggah olehosama
- Optimization of Scheduling in FMS using Heuristic Approaches: A Case StudyDiunggah olehEditor IJTSRD
- Use of Genetic Algorithm to Resolve the Problem of Finite State Machine ConstructionDiunggah olehDevendra Sharma
- Survey Table CorrectedDiunggah olehAnonymous TxPyX8c

- Diaspora and Memory Figures of DisplacemDiunggah olehHayv Kahraman
- my mission as a teacher can be summed up in three partsDiunggah olehapi-383969732
- Ethic Definition PDFDiunggah olehMariah
- Bio Psychology Lecture Notes Chapter 6 the Visual SystemDiunggah olehGeneric_Persona
- selectedthesissectionsalstonDiunggah olehapi-349198821
- learning segment-ela 2Diunggah olehapi-254944388
- Assembling bodies_KayAre.pdfDiunggah olehKay Are
- MODEL BASED ANALYSIS OF TEMPERATURE PROCESS UNDER VARIOUS CONTROL STRATEGIES | J4RV3I12010Diunggah olehJournal 4 Research
- 3. Role of the Nurses as Health EducatorDiunggah olehJuli-anne Villarico Andres
- Chapter 6Diunggah olehAditi Gupta
- Optimism and Entrepreneurship - A Double-Edged SwordDiunggah olehNesta
- Boeckx (2008) - Islands.pdfDiunggah olehpalingenesia
- cando namechartsDiunggah olehapi-322200151
- march 13 lesson plan r+jDiunggah olehapi-283068954
- 373Diunggah olehRamakrishnan Harihara Venkatasubramania
- 182329726-Goldman-Kant-Heidegger-and-the-Circularity-of-Transcendental-Inquiry-pdf.pdfDiunggah olehAngela Dean
- Appelbaum - Performance AppraisalDiunggah olehmametosaurus
- edu 5230Diunggah olehapi-438490403
- Integrating Sense of Humor as a Teaching Strategy among Students of LPU-BatangasDiunggah olehAsia Pacific Journal of Multidisciplinary Research
- BAI 2Diunggah olehbingo080104
- p&p1-idrusDiunggah olehAhmad Lutz
- The Noesis and NoemaDiunggah olehAquamarine Ruby Goldmagenta
- Sambahsa Reference DocumentDiunggah olehcafaristeir
- Factors Influencing Low English Fluency Level in Tadika of the Rural Areas by Balvinder KaurDiunggah olehAin Nor Fatina
- One Week Lesson Plan.templateDiunggah olehAmine El Kharaze
- A Manual of Modern Scholastic Philosophy v2 1000059811Diunggah olehDelicatvs
- Tell-Tale Heart.pdfDiunggah olehAncuta Inna
- tp4Diunggah olehapi-282173119
- Experimental Research Method PDFDiunggah olehLindsey
- Roles of TechnologyDiunggah olehRachel Navarez